`dynesty_runplot`

Draw the standard nested-sampling diagnostic panels for a dynesty run: live points, log-likelihood, importance weight, and accumulated evidence (ln Z) versus iteration.

Based on: dynesty's run-plot diagnostics (no Matplotlib Axes equivalent). This is a specialist method — it fills a multi-panel figure rather than drawing one primitive.

Axes: a multi-panel style card (panels ax0, ax1, ax2, ax3); the a4paper_2x1 / dynesty_runplot card is intended for it.

Data

The layer's data source must contain the run trace columns dynesty produces (live points, log-likelihood, log-weight, log-evidence over iterations). No coordinates block is needed — the method reads the trace itself.

Style

Key	Default	Purpose
`panels`	`[nlive, likelihood, importance, evidence]`	Which panels to draw
`axes`	`[ax0, ax1, ax2, ax3]`	Panel axis ids to draw into
`logplot`	`false`	Log-scale the y axes
`kde`	`true`	Smooth the importance-weight panel with a KDE
`nkde`	`1000`	KDE sample count
`lnz_error`	`true`	Shade the `ln Z` uncertainty band
`lnz_error_levels`	—	Error-band σ levels
`max_x_ticks` / `max_y_ticks`	`8` / `3`	Tick density
`columns`	—	Map trace column names if they differ from the defaults

Example

- name: nested_diagnostics
  enable: true
  style: [a4paper_2x1, dynesty_runplot]
  layers:
    - name: runplot
      data:
        - source: dynesty_trace
      method: dynesty_runplot
      style:
        logplot: false
        kde: true

Notes

This method is for inspecting sampler convergence, not for physics result figures.
Use the columns style mapping if your trace columns are named differently from dynesty's defaults.